状态和输入受限的切换奇异布尔控制网络的最优控制

本文研究了状态和输入均受限的切换奇异布尔控制网络的最优控制问题.利用矩阵半张量积方法获得受限切换奇异布尔控制网络的等价代数形式.然后通过类似针变化得到了存在最优控制的必要条件,并且提出了一个算法设计切换序列和控制策略使收益函数最大化.最后给出例子验证所得结果的有效性.     

布尔控制网络的集成集可控

研究一类布尔控制网络的集成集可控性和集成集镇定性问题.首先,利用矩阵的半张量积理论,给出布尔控制网络等价的代数表示;其次,通过自由控制序列研究布尔控制网络的集成集可控性,并给出相应的充分必要条件,对于布尔控制网络的集成集镇定性问题,使其转换为集成集可控性问题,并给出相应的判定定理;最后,给出数值例子说明所提出方法的有效...     

基于事件触发和状态翻转的布尔控制网络输出跟踪

利用矩阵半张量积研究事件触发和翻转控制共同作用下布尔控制网络的输出跟踪问题.首先,基于布尔控制网络代数状态空间表示,构造增广系统将输出跟踪问题转化为状态集镇定问题;其次,得到布尔控制网络在两种控制下输出跟踪问题有解的充要条件,并在满足该条件时提出一种基于最小翻转节点集时间最优控制设计方法,进一步给出有限时间内寻找翻转节点集的计算过程;最后,给出一个算例说明结果的可行性.     

受扰布尔控制网络全局可重构分析与状态估计

受扰布尔控制网络的状态转移,因受未知干扰影响而具有不确定性,这对状态观测器设计带来了困难.本文主要研究了受扰布尔控制网络全局可重构性问题,并在此基础上设计状态观测器.首先,将受扰布尔控制模型转化为多个子系统的切换未知布尔控制网络模型,在此基础上,提出了受扰布尔控制网络的4种不同状态集.其次,基于状态集估计方法,对受扰布尔控制网络状态估计问题进行分析.再次,提出有限时间可重构与全局可重构性概念;同时,根据状态集估计与状态转移分析,分别给出有限时间可重构判定算法与全局可重构性证明的充要条件.最后,给出观测器设计方法,并通过例子证明了本文提出方法的可行性.     

受限布尔网络发展现状

布尔网络可以简洁有效地描述作用在有限集上的动态离散模型.然而,随着研究的深入以及一些实际问题的需要,传统的布尔网络已经不能满足建模的需求,由此衍生出受限布尔网络,通过矩阵半张量积,该类型的网络可以转化为便于处理的等价代数表示.鉴于此,对受限布尔(控制)网络的来源、受限形式和相关问题,作了概括与总结.对于受限布尔网络中出现的典型问题、规范化与可解性,理清了其发展脉络与研究现状;对受限布尔网络的拓扑结构整理了相关结果.另一方面,在受限布尔控制网络部分,着重总结其能控性的发展现状,将现有的能控性分析方法归为Dimitriy-Michael方法和预反馈方法两大类,并分别介绍其分析过程.总结受限布尔控制网络在设计能控、镇定、最优控制信号等问题中的一些常用方法(输入-状态关联矩阵方法和Floyd算法),以及牵引控制和干扰解耦等其他研究方向.     

布尔控制网络的输出稳定与镇定

矩阵的半张量积是将逻辑变量转化为向量研究的主要工具.本文利用半张量积把逻辑控制系统表示为离散时间仿射线性系统,在逻辑系统的状态空间框架下研究了以布尔控制网络为代表的逻辑动态系统的输出稳定与镇定.首先给出布尔网络输出稳定的定义,研究了布尔网络输出稳定的充要条件;其次讨论了布尔控制网络的输出镇定,分别得到了布尔控制网络由常值输入变量、自由控制序列、状态反馈控制序列输出镇定的条件.本文讨论的系统输出稳定与镇定是(部分)变量稳定与镇定的推广.     

LFNDIT:从不确定状态变换学习布尔网络

布尔网络是一种重要的基因调控数学模型,从布尔网络的状态变换推断其结构以发现基因之间的调控关系是布尔网络研究中长期关注的重要问题。已有的归纳逻辑程序算法不能从布尔网络的不确定(解释)状态变换学习推断其网络结构。为此,文中提出了非确定解释转换学习(Learning From Non-deterministic interpretation Transitions,LFNDIT)算法从布尔网络异步更新语义下的解释变换学习其网络结构。首先将异步更新语义下的不确定解释变换集转换成确定解释变换集,然后利用Inoue等提出的从1步解释转换学习(Learning From 1-step state transition,LF1T)算法计算其对应的正规逻辑程序(布尔网络)。该算法的完备性得到了证明,初步的实验结果表明,该方法能有效地从不确定状态变换计算布尔网络的结构,从而为发现布尔网络的结构提供了新的思路。     

优先状态估计的双深度Q网络

深度强化学习探索问题中,需要根据环境给予的外部奖赏以作出决策,而在稀疏奖赏环境下,训练初期将获取不到任何信息,且在训练后期难以动态地结合已获得的信息对探索策略进行调整.为缓解这个问题,提出优先状态估计方法,在对状态进行访问时给予优先值,结合外部奖赏一并存入经验池中,引导探索的策略方向.结合DDQN(Double Dee...     

基于牵制控制的切换布尔网络的分布式集合镇定

切换布尔网络是一种典型的网络化控制系统, 在基因调控、信息安全、人工智能、电路设计等领域具有重 要应用. 本文基于牵制控制方法, 研究切换布尔网络在任意切换下的分布式集合镇定问题. 首先, 利用矩阵半张量积 方法,得到切换布尔网络的代数形式. 其次, 基于代数形式, 提出构造性的算法来实现切换布尔网络在牵制控制的 作用下任意切换集合镇定, 并设计出状态反馈牵制控制器. 再次, 利用逻辑矩阵分解技术和分布式控制方法, 设计任 意切换下切换布尔网络的分布式集合镇定控制器, 并提出分布式控制器存在的充分条件. 文中给出3个例子来说明 所获得结果的有效性.     

基于Q学习算法和BP神经网络的倒立摆控制

Q学习是Watkins[1]提出的求解信息不完全马尔可夫决策问题的一种强化学习方法.将Q学习算法和BP神经网络有效结合,实现了状态未离散化的倒立摆的无模型学习控制.仿真表明:该方法不仅能成功解决确定和随机倒立摆模型的平衡控制,而且和Anderson[2]的AHC(Adaptive Heuristic Critic)等方法相比,具有更好的学习效果.     

Controllability and stabilization of Boolean control networks by the auxiliary function of flipping

In this article, a hybrid method combining flip matrix approach and an open (or closed) loop control is proposed to study global controllability and stabilization of Boolean control networks (BCNs). First, the necessary and sufficient condition for global controllability of BCNs, by flipping some members of a perturbation set and under free control sequences, is proposed. After that, using a search algorithm, the minimal perturbation sets for global controllability (MS‐GCs) of the BCN are obtained. Next, we propose a necessary and sufficient criterion for global stabilization of BCNs by flipping some members of a perturbation set and under a state feedback control. Similarly, an algorithm is given to search for the minimal perturbation sets for global stabilization (MS‐GSs) of the BCN. Moreover, the time‐optimal MS‐GSs of the BCN are also obtained by an algorithm. Some examples are given to illustrate the effectiveness of the results.     

概率布尔控制网络的集可控

本文研究概率布尔控制网络的集可控性问题.首先,利用矩阵半张量积方法,得到概率布尔控制网络的代数表示.其次,借助一个新的算子构造不同的可控矩阵,进而通过可控矩阵考虑自由控制序列和网络输入控制下概率布尔控制网络的集可控性问题,得到了概率布尔控制网络集可控性的充要条件.最后,给出数值例子说明本文结果的有效性.     

Reconstructibility analysis and set-observer design for Boolean control networks

This paper deals with the issues of reconstructibility analysis and set-observer design for Boolean control networks (BCNs). A state estimation set is considered to find the state estimation of BCNs, and a reconstructibility judgment matrix (RJM) is obtained by compact vector of state estimation set. Then, a kind of state-reconstructible tree is constructed, and algorithms are also provided to construct it. The state-reconstructible tree can categorize BCNs to be globally reconstructible, locally reconstructible, and unreconstructible. The sufficient and necessary conditions for the reconstructibility of BCNs are given. Meanwhile, the relationship between input–output trajectory and estimated state is derived. Next, with the knowledge of input–output sequence, a set-observer is designed such that the state of BCNs for any global reconstructible BCNs is uniquely determined in a finite time. Some examples are also given to illustrate the effectiveness of the proposed methods.     

On controllability and stabilizability of probabilistic Boolean control networks

The controllability of probabilistic Boolean control networks(PBCNs)is first considered.Using the input-state incidence matrices of all models,we propose a reachability matrix to characterize the joint reachability.Then we prove that the joint reachability and the controllability of PBCNs are equivalent,which leads to a necessary and sufcient condition of the controllability.Then,the result of controllability is used to investigate the stability of probabilistic Boolean networks(PBNs)and the stabilization of PBCNs.A necessary and sufcient condition for the stability of PBNs is obtained first.By introducing the control-fixed point of Boolean control networks(BCNs),the stability condition has finally been developed into a necessary and sufcient condition of the stabilization of PBCNs.Both necessary and sufcient conditions for controllability and stabilizability are based on reachability matrix,which are easily computable.Hence the two necessary and sufcient conditions are straightforward verifiable.Numerical examples are provided from case to case to demonstrate the corresponding theoretical results.     

Finite-time observability of probabilistic Boolean control networks

In this paper, the finite-time observability (FO ) of probabilistic Boolean control networks (PBCNs ) based on set reachability and parallel extension is investigated. Under algebraic state space representation of PBCNs , finite-time set reachability (FSR ), finite-time single input sequence set reachability (FSSSR ), finite-time arbitrary input sequence set reachability (FASSR ), as well as finite-time output feedback set reachability (FOSR ) are addressed, where some efficient criteria are proposed. In the second part of this paper, several types of FO are provided, which can be converted to the corresponding FSR based on the parallel extension technique. Consequently, the obtained results for FSR are utilized to deal with FO . Finally, two examples including a biological network are given to demonstrate the theoretical results.     

布尔控制网络的能控性与能观性

利用矩阵的半张量积,布尔控制网络被转化为离散时间系统.本文从离散时间系统的结构矩阵出发,讨论了逻辑控制系统的能控能观性条件,得到了一个新的能控性条件.新的条件简化了原有能控性矩阵的计算复杂性,矩阵的最高阶数由原来的2m+n降到了2n.另外,还得到了检验布尔控制网络能观性的条件.与原有条件相比,新的条件更容易计算检验.最后,给出一个实例,检验给出的能控能观性判断条件的正确性.     

Simultaneous stabilization for a set of Boolean control networks

This paper investigates simultaneous stabilization of a collection of Boolean control networks (BCNs) by using the semi-tensor product method, and presents a number of new results. First, an algebraic expression of the BCNs is obtained by the semi-tensor product, based on which some necessary and sufficient conditions are presented to solve the simultaneous stabilization problem by a free control sequence, a state-feedback control, and an output-feedback control, respectively. Second, using the column stacking form of matrices, a new procedure is established to design both state-feedback and output-feedback controllers for the simultaneous stabilization problem. The study of two illustrative examples shows that the new results obtained in this paper are very effective in solving simultaneous stabilization of a collection of BCNs.     

Feedback control design for the complete synchronisation of two coupled Boolean networks

In the literatures, to design state feedback controllers to make the response Boolean network synchronise with the drive Boolean network is rarely considered. Motivated by this, feedback control design for the complete synchronisation of two coupled Boolean networks is investigated in this paper. A necessary condition for the existence of a state feedback controller achieving the complete synchronisation is established first. Then, based on the necessary condition, the feedback control law is proposed. Finally, an example is worked out to illustrate the proposed design procedure.